When I calls them, I really calls them …

See this bit of propaganda
and the money quotes on Linux within
.

Q: A couple of you have brought up Linux. Based on your experience, how does a Windows-based HPC platform compare to Linux-based HPC clusters in areas like development, setup, maintenance, interoperability, scalability and applications?

Wortman: The setup and management of Windows Compute Cluster Server vs. a Linux cluster is worlds apart for us. The proof of this is the fact that a Windows technician with no HPC experience can set up a cluster from scratch in a matter of hours. Linux clusters simply take more care and feeding, and substantial knowledge of Linux, which nearly all biomedical researchers lack.

Let me get this straight … I want to make sure I understand this … You really want your expensive HPC computing resource/platform setup by someone who doesn’t know HPC?

Sort of like you wanting that expensive database server set up by a windows technician with no knowledge of databases? Or that nice web server set up by someone with no knowledge of web servers. Or …

The point should be obvious.

There were other bits about “linux clusters being hard to set up and then appropriately tune up”. Ummm… I guess they didn’t see this.

It is not hard to cable boxes together, and get something running quickly. For the simplest cluster designs, this works very well with Rocks. You answer a few questions after booting from a DVD, and your system is up pretty quickly afterwords. Clusters from 1 to 2048 nodes. Doesn’t matter, the admin time does not scale with the number of nodes.

Of course for more complex computational analysis you need better cluster designs. It is beyond me how a windows technician with no knowledge of HPC could possibly be an asset in this case.

While I am sure the windows technician could in fact install windows, I would be hardpressed to imagine they could deal with tuning for bursty file server workloads, performance optimization under a variety of run circumstances, or even deal with ameliorating problems with the same.

I am not sure what to think about anyone willing to trust the setup of a critical component to someone who doesn’t quite grasp what the component does, how it all works … Yeah, you can tinker with your car’s engine. You might even be able to fix things if you are not trained to, by carefully following a set of instructions. But there is a very limited range of things you can do, and you would lack the experience to deal with issues that often do arise. Thats why, unless you are experienced with dealing with your cars engine, you ask it to be serviced by someone who knows how the engine works, and has a fighting chance of identifying and fixing problems. Sure you will save a little money upfront if you bring your car to your cosmotologist to get it fixed, but there is a pretty good chance that they can do a very limited range of operations before they are completely lost. Same with the windows technician.

Astounding.

But I called it, and I was spot on. Windows Cluster is part of tactical efforts of Microsoft to deal with Linux as a threat. Not part of a strategy to grow in the growing HPC market. HPC has been growing at a tremendous clip long before Microsoft offered any products here.

Viewed 10214 times by 1820 viewers

4 thoughts on “When I calls them, I really calls them …

  1. I can’t decide if I want to get into this again, because it’s hard for me to parse your last paragraph. (Yes, Microsoft is employing tactical efforts to deal with Linux as a threat. Yes, they want to see the HPC market grow, and they want to grow their market share as well. I don’t think those things are mutually exclusive.)

    But I think it’s worth pointing out that the quote you pulled is NOT from a Microsoft shill; according to that article, Matt Wortman is the director of computational biology and IT at the Genome Research Institute at the University of Cincinnati. I imagine he has at least some experience with clusters, and with managing people who run clusters. Your treatment of his quote makes it appear as if he’s a know-nothing PR flack. His opinion carries some weight.

  2. Hmmm. Not sure if I am going to get into this past this message. I don’t appreciate when people twist meaning to come up with one of their own construction which they subsequently attribute to me.

    I am glad you agree that this is a tactical move on their part. Some people (yourself as I remember) took issue with that previously. We had long discussions of this.

    The cluster market was growing nicely before Microsoft arrived, and it appears this growth is not slowing, but accelerating.

    Microsoft is seeking to create its own environment, and is doing so outside of the current one. Not together with the existing market to grow a larger version of the current market, but competitive with the current market. Their choice.

    I disagree with the concept of having “a technician … with no HPC experience” set up a complex HPC environment. I described why I took issue with it using the example of fixing ones car, and tried a little humor with the cosmotology bit. That may have escaped some readers.

    I did take issue with his statement. I think a quite a number of people would have issues with his comment, assuming that it is not PR filtered.

    Matt Wortman is a respected scientist, working with good groups at OSC and elsewhere. Don’t try to stick words in my mouth and then accuse me of saying or implying something I did not. Thats rather disingenuous.

  3. You called his quote “propaganda” in the first sentence of your post, Joe. I didn’t twist your words. I said “Your treatment of his quote makes it appear as if he???s a know-nothing PR flack.”

    You’re right, you didn’t call him a “know-nothing PR flack,” and I apologize if it seemed I accused you of using those words.

    But you did accuse him of propagandizing, and your condescending words (“Let me get this straight ??? I want to make sure I understand this ???” “Ummm??? I guess they didn???t see this”) sure didn’t seem to lend the respect that Dr. Wortman surely deserves.

    You, too, deserve respect, btw. You are definitely respected in the HPC community (which is more than I can say for myself!). And it just goes to show you: smart, respected people have different opinions about the state of clustering.

Comments are closed.